Joint Angle Drift Remedy Of PA10 Robot Arm At Joint-Acceleration Level

Redundant manipulators are robots having more degrees of freedom (DOF) than the minimum number to perform a given end-effector primary task. In this sense, the redundancy of joint motion complicates the manipulator control problem. One fundamental issue in operating such systems is the redundancy-resolution problem. As we know, multiple solutions (or even an infinite number of solutions) may exist for a redundant manipulator performing a prescribed end-effector task. The conventional solution to such a redundant-resolution problem (especially when resolved at joint-velocity level) is the pseudoinverse-based formulation; i.e., one minimum-norm particular solution plus a homogeneous solution. The research of recent ten years shows that the redundancy-resolution problem could be solved through a more favorable manner via online optimization techniques..Joint angle drift problem is that when a manipulator's end-effector traces a closed path in its workspace, the joint angles may not return to their initial ones after completing such an end-effector task. This is also called non-repetitive problem. The conventional pseudoinverse-based solution also may not be repetitive. The previous research of repetitive motion planning are mainly at the joint-velocity level, sometimes, this may not be applicable to some manipulators which are controlled by acceleration or torque. Furthermore, the joint acceleration limits of manipulators can not be constrained in the joint-velocity-level scheme. In this paper, a quadratic-programming (QP) based method is employed for joint angle drift problem of redundant robot manipulators. In addition, the physical constraints such as joint limits, joint velocity limits and joint acceleration limits are incorporated into the problem formulation of such a redundancy-resolution scheme. The scheme is finally reformulated as a strictly-convex quadratic-programming (QP) problem and resolved at joint-velocity level.To do this, a simplified LVI-based primal-dual neural network is presented for the online repetitive motion planning. To demonstrate the effectiveness of such a QP based method, PA10 robot manipulator are tested to perform different kinds of end-effector trajectories through computer simulation.